return NotImplemted for unsupported operations in __array_ufunc__ and implement matmul

test/test_value_array.py

@@ -293,3 +293,12 @@ def test_unique() -> None:
     unit = tu.MHz
     v_arr = xs * unit
     assert np.array_equal(v_arr.unique(), np.unique(xs) * unit)
+
+
+def test_matmul() -> None:
+    a = np.random.random((3, 4))
+    b = np.random.random((4, 3)) * tu.ns
+    c: tu.TimeArray = a @ b  # type: ignore[assignment]
+    d: tu.TimeArray = b @ a  # type: ignore[assignment]
+    assert c.allclose((a @ b[tu.us]) * tu.us)
+    assert d.allclose((b[tu.s] @ a) * tu.s)

tunits/core/cython/with_unit_value_array.pyx

@@ -116,6 +116,10 @@ class ValueArray(WithUnit):
                 return self ** 2
             if ufunc == np.reciprocal:
                 return self.__rtruediv__(1)
+            if ufunc == np.matmul:
+                if isinstance(inputs[0], ValueArray):
+                    return inputs[0].__matmul__(inputs[1])
+                return inputs[1].__rmatmul__(inputs[0])
 
         if ufunc in [
             np.greater,
@@ -133,7 +137,7 @@ class ValueArray(WithUnit):
         if self._is_dimensionless():
             return getattr(ufunc, method)(*(np.asarray(x) for x in inputs), **kwargs)
 
-        raise NotImplemented
+        return NotImplemented
 
     @property
     def dtype(WithUnit self) -> np.dtype:
@@ -160,4 +164,11 @@ class ValueArray(WithUnit):
     def to_proto(self, msg: Optional['tunits_pb2.ValueArray'] = None) -> 'tunits_pb2.ValueArray':
         ret = _ndarray_to_proto(self.value, msg)
         ret.units.extend(_units_to_proto(self.display_units))
-        return ret
+        return ret
+
+    def __matmul__(WithUnit self, other: np.ndarray):
+        return self.__with_value(self.value @ other)
+
+
+    def __rmatmul__(WithUnit self, other: np.ndarray):
+        return self.__with_value(other @ self.value)